A Landsat-based atlas of monthly burned area for Portugal, 1984–2021

Abstract

Portugal is the most fire-prone country in Southern Europe. Time series of satellite imagery have great potential to improve fire mapping and explore the spatiotemporal trends of burned areas. Thus, the objectives of this work were to disaggregate the Portuguese Annual Fire Atlas burned area patches into individual events according to their date of occurrence estimated from Landsat temporal series, and to assign them the closest date (day and month) of the detected change. From the resulting Monthly Fire Atlas, we assessed intra-annual fire variability in Portugal. Our analysis was performed from 1984 to 2021 for the entire mainland Portugal. We selected the closest fire date based on the index of the lowest ΔNBR in the time series. Historically, most of the large fires were concentrated in the summer season. However, when daily mean areas are analyzed, peaks can be identified due to fire events that occurred at a time of the year when burned areas are typically much lower. The results were validated using MODIS and VIIRS active fires. Accuracies of individual years fluctuated according to the satellite and cloudiness of each year. For the entire 37 analyzed years, when considering 16 and 32 days of time gap, the Monthly Fire Atlas achieved 84% and 91% of accuracy, respectively.

Keywords

• Fire mapping
• Landsat image time series
• Google Earth Engine